FIGURE 4.

Performance of a DCR-based Convolutional Neural Network (CNN) classifier for predicting scoliosis-associated genotypes of Fibrillin-1 (FBN1). (A) Training loss per training epoch was plotted as a boxplot for all loss values per epoch, with the average value per epoch curved. (B, C) Comparison of the PCA1 (B) and PCA2 (B) reduced from the fully connected layer of the trained CNN classifier. ROC_AUC (D), Confusion matrix (E) by the trained CNN classifier based on independently sampled variants. Benign: low scoliosis risk, Pathogenic: high scoliosis risk. ****p < 0.0001, ns: no significance.